RESEARCH ARTICLE Open Access

Dynamics of toxigenic Clostridiumperfringens colonisation in a cohort ofprematurely born neonatal infantsAlexander G. Shaw1*† , Emma Cornwell2†, Kathleen Sim2, Hannah Thrower2, Hannah Scott2, Joseph C. S. Brown3,Ronald A. Dixon3 and J. Simon Kroll2

Abstract

Background: Clostridium perfringens forms part of the human gut microbiota and has been associated with life-threatening necrotising enterocolitis (NEC) in premature infants. Whether specific toxigenic strains are responsible isunknown, as is the extent of diversity of strains in healthy premature babies. We investigated the C. perfringenscarrier status of premature infants in the neonatal intensive care unit, factors influence this status, and the toxicpotential of the strains.

Methods: C. perfringens was isolated by culture from faecal samples from 333 infants and their toxin gene profilesanalysed by PCR. A survival analysis was used to identify factors affecting probability of carriage. Competitivegrowth experiments were used to explore the results of the survival analysis.

Results: 29.4% of infants were colonized with C. perfringens before they left hospital. Three factors were inverselyassociated with probability of carriage: increased duration of maternal milk feeds, CPAP oxygen treatment andantibiotic treatment. C. perfringens grew poorly in breast milk and was significantly outperformed by Bifidobacteriuminfantis, whether grown together or separately. Toxin gene screening revealed that infants carried isolates positivefor collagenase, perfringolysin O, beta 2, beta, becA/B, netB and enterotoxin toxin genes, yet none were observed tobe associated with the development of NEC.

Conclusions: Approximately a third of preterm infants are colonised 3 weeks after birth with toxin gene-carrying C.perfringens. We speculate that increased maternal breast milk, oxygen and antibiotic treatment creates anenvironment in the gut hostile to growth of C. perfringens. Whilst potentially toxigenic C. perfringens isolates werefrequent, no toxin type was associated with NEC.

Trial registration: clinicaltrials.gov NCT01102738, registered 13th April 2010.

Keywords: Clostridium perfringens, Breast milk, Toxins, Necrotising enterocolitis

BackgroundClostridium perfringens is an archetypal pathosymbiont,forming part of the gut commensal microbiota inhumans and animals, but also capable of producing dev-astating disease by way of its toxin arsenal. This anaer-obic, Gram-positive spore-former is the leading cause oftraumatic gas gangrene in humans [1] and one of the

most common causes of food poisoning, responsible foran estimated 1 million cases in the US each year [2]. C.perfringens has been linked to necrotizing enterocolitis(NEC) – an inflammatory bowel disease with high mor-tality – in preterm neonates [3–7]. This condition hasseveral predisposing factors; an immature gut, the pres-ence of bacteria and potentially hypoxia and ischemiaaround the site of pathology [3]. A single causative or-ganism has however remained elusive, with studies hav-ing found associations with other components of the gutmicrobiota such as the Enterobacteriaceae [7–9]. Thetoxic potential of C. perfringens in relation to the genesis

© The Author(s). 2020 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: a.shaw@imperial.ac.uk†Alexander G. Shaw and Emma Cornwell contributed equally to this work.1Department of Infectious Disease Epidemiology, Imperial College London,London, UKFull list of author information is available at the end of the article

Shaw et al. BMC Pediatrics (2020) 20:75 https://doi.org/10.1186/s12887-020-1976-7

of NEC remains intriguing. A role has also been pro-posed for Clostridium spp. in infant atopy and allergicsensitisation [10, 11] and autism [12, 13].Previous studies investigating clostridial carriage in ap-

parently healthy premature neonates have often involvedsmall sample sizes, [14, 15] or reflected the opportunityto study control groups recruited for studies of topicssuch as the effect of different feeding or probiotic re-gimes [16]. We have recruited a large cohort of well-characterised premature neonates who were treated ac-cording to standard care protocols and provided faecalsamples from birth until their departure from hospital.The microbiota profiles of these neonates were previ-ously characterised by next-generation sequencing whichidentified significantly higher levels of C. perfringens in asubset of infants who were diagnosed with NEC com-pared to controls [7]. In the present current study, ourfirst aim has been to use a culture-dependent approachto build up a collection of neonatal C. perfringens iso-lates and establish colonisation dynamics in the cohort.We have gone on to relate this to perinatal factors whichwe hypothesise will influence rates of carriage.We have investigated the toxin gene carriage rates in

these neonatal C. perfringens isolates. C. perfringensstrains encode a formidable arsenal of more than 20toxins [17, 18]- many with cytotoxic effects - includingthe well-characterised alpha toxin, beta toxin, and the C.perfringens enterotoxin (CPE, encoded by the cpe gene),which are respectively implicated in the human diseasesgas gangrene, necrotic enteritis (Pigbel) and food poison-ing (see Supplementary Material 1). Previous studieshave shown that toxigenic C. perfringens strains arepresent in healthy adult populations (see SupplementaryMaterial 1), but there is a lack of data for neonatal popu-lations and also for toxins other than the 4 major and 2minor toxins that make up standard multiplex PCRscreens for assigning toxin type and investigating casesof food poisoning [19]. We here report the result ofscreening for the presence of 11 toxin genes; cpa, cpb,etx, itx, cpe, cpb2, netB, becA, becB, pfoA and colA (seeSupplementary Material 1 for characterisations) in ourneonatal isolates, seeking possible associations with NECand establishing the normal clostridial toxin gene car-riage rates in premature infants during their stay on theneonatal intensive care unit.

MethodsStudy populationInfants born < 32 weeks gestation admitted to an Imper-ial College Healthcare National Health Service Trustneonatal intensive care unit (St Mary’s Hospital, QueensCharlotte’s and Chelsea Hospital) between January 2010and December 2011 were eligible for inclusion in ourecological study “Defining the Intestinal Microbiota in

Premature Infants” (The Neonatal Microbiota (NeoM)Study). Both hospitals have identical antibiotic and feed-ing protocols and staff members rotate between sites.Detailed daily clinical records were collected for all par-ticipants. Of the 369 babies recruited, 333 provided fae-cal samples for analysis in the present study (a total of1399 samples).

Sample collectionWe collected every faecal sample produced by partici-pants between admission and discharge. Samples werecollected by nursing staff from diapers using a sterilespatula and placed in a sterile DNAase-, RNAase-freeEppendorf tube. These were stored at − 20 °C within 2 hof collection and were transferred to − 80 °C storagewithin 5 days. Approximately one sample per week wasselected for culture.

Clostridium culture protocolSelective culture for clostridia species was performedusing an ethanol shock methodology to eliminate non-spore forming organisms [20]. 25 mg of faeces was addedto 500 μl of cooked meat broth (Oxoid) and 500 μl of100% ethanol and vortexed for 10 sec before incubatingfor 30 min at room temperature. A sterile loop wasdipped in the supernatant and streaked onto a fastidiousanaerobic agar plate supplemented with 0.1% sodiumtaurocholate hydrate (Sigma) and 5% defibrinated sheepblood (Oxoid). Plates were incubated anaerobically at37 °C for 48 h in an Oxoid 2.5 l AnaeroJar plus Anaero-Gen sachet. Resulting colonies were examined and foreach morphologically distinct isolate, four colonies weresubcultured by streaking onto Colombia agar with sheepblood (Oxoid) plates which were incubated anaerobicallyat 37 °C for 48 h. Single colonies were used for identifi-cation to the species level by matrix-assisted laser de-sorption/ ionization–time of flight (MALDI-TOF) usinga Bruker Microflex LT (Bruker Daltonics), and the re-mainder of the subculture stored in 70% brain heart in-fusion (BHI) broth (Oxoid), 30% glycerol (Sigma) at −80 °C.

Toxin typing of Clostridium perfringens by polymerasechain reaction (PCR)PCR reactions were performed to detect cpa, cpb, iA, etx,cpe and cpb2 (multiplex PCR), becA and becB (duplexPCR) and netB, pfoA, and colA (single PCR) and visualisedby gel electrophoresis. For details see SupplementaryMaterial 2.

Breast milk growth competition experimentBifidobacterium infantis (NCIMB 702255 – isolatedfrom infant intestine), and Clostridium perfringens iso-lated from the faeces of an infant enrolled in our study

Shaw et al. BMC Pediatrics (2020) 20:75 Page 2 of 11

were cultured either in Wilkins-Chalgren broth (WCB)(Oxoid) supplemented with 5 g/l soya peptone (Oxoid),or human breast milk. Three donors gave breast milk onthree occasions, for which informed consent was ob-tained and standardised expression protocols used.Breast milk was expressed on average 16 h before eachexperiment, was kept refrigerated at all times, and had acommensal microbial load (anaerobic) below the NICE-recommended limit (105 colony forming units (CFU)/mltotal viable organisms or 104 CFU/ml Enterobacteriaceaeor Staphylococcus aureus) [21], which was calculated byplating directly onto Colombia agar with sheep blood(Oxoid) and incubating for 48 h at 37 °C, followed byspecies identification using MALDI-TOF. The majorityof organisms found in this manner were Staphylococcusand Actinomyces spps. Staphylococcus spp. were grownin every breast milk sample, whilst Actinomyces spp.were found in the milk of only one mother (see Supple-mentary Material 3). For bacterial counts, C. perfringenswas grown on Tryptose Sulphite Cycloserine (TSC) agarplates (Oxoid) [22], and incubated for 24 h at 37 °C, andB. infantis on Bifidus Selective Medium (BSM) agarplates (Sigma) incubated for 48 h at 37 °C. All cultivationwas performed under anaerobic conditions using anOxoid 2.5 l AnaeroJar plus AnaeroGen sachet in a shak-ing or static 37 °C incubator.Bacteria were grown in WCB to an optical density of

approximately 0.7, and approximately 1 × 104 colonyforming units (CFUs) added to 2.5 ml of breast milk or2.5 ml of WCB in 15 ml bioreactor tubes (TTP). Mediawere inoculated with C. perfringens or B. infantis (mono-culture), or both (co-culture). C. perfringens and B.infantis were enumerated after 20 h incubation by plat-ing dilutions in triplicate onto TSC agar and BSM agarrespectively. Breast milk from three different donors wasused and experiments carried out in triplicate (to totalnine experiments).

StatisticsStatistical analyses were performed in the R statisticalpackage (version 3.3.1). A Cox Proportional-Hazard re-gression function from the ‘survival’ package was usedfor the survival analysis. An initial univariate analysiswas performed to determine any significant relation be-tween each clinical factor in Table 1 and C. perfringenscolonisation (defined as the earliest faecal sample whereC. perfringens was isolated). Factors found to remain sig-nificant after a multiple hypothesis (Bonferroni) correc-tion were entered into an iterative multivariate model,with factors found to have a p value of > 0.1 being re-moved at each iterative step. Barnard’s test was per-formed to detect associations between antibiotic use andbreast milk feeds and long-term carriage, and betweentoxin genes and NEC development. Growth rate

comparisons were performed using generalized linearmodels and the ‘mass’ package using both the meanfold-change for each experimental triplicate and themean absolute change in CFU. P-values shown were de-rived from the fold-change models, with similar resultsby either metric.

ResultsC. perfringens incidence and clinical factors associatedwith its colonisationFaecal samples and complete clinical notes were avail-able for 333 infants. The demographics of these babiesare shown in Table 1. C. perfringens was isolated in fae-cal samples from 98 of the infants (29.4%).Colonisation data were used to predict the risk of col-

onisation of the infant gut by C. perfringens over timethrough a survival analysis (Fig. 1). By the day of life ofthe median stay on the neonatal unit for the cohort (29days), a predicted 36% of infants would be colonised(95% confidence band 25, 43%).We next repeated the survival analysis with the aim of

determining clinical variables (shown in Table 1) thatwere associated with C. perfringens colonisation. A seriesof univariate models were created and significantly asso-ciated variables (after multiple hypothesis correction) areshown in Table 2.For each significant factor, values are provided at

each of its quartiles to illustrate the spread of data.The exponentiated coefficients provide the change inrisk of colonisation per unit of each factor. Riskchange indicates the relative change in risk of colon-isation between the minimum value (0%) and the 75%quartile for a given clinical factor. Abbreviations: CI,confidence interval.Given the potential for correlation between these

variables, a multivariate survival analysis was used toidentify a minimal set of clinical factors to best pre-dict colonisation. Four factors were found to remainsignificant in this model, with associations betweenincreased probability of C. perfringens colonisationand fewer days of CPAP with supplemental oxygen(CPAP oxygen), fewer days of maternal milk feeds(via feeding tube), fewer days of breast feeding andfewer days of antibiotics over the course of the in-fant’s admission. The variation of colonisation prob-abilities when the infant cohort is divided intoquartiles for each of these factors is shown in Fig. 2.As the risk of colonisation for each infant is associated

with the combined effects of each of these factors, weperformed a multivariate analysis where each infant inthe cohort was annotated either “low” (<= median) or“high” (>median) for each of the four factors. Multiplesets of analyses were run, illustrating the effects ofhigher than average measure of a single factor, or

Shaw et al. BMC Pediatrics (2020) 20:75 Page 3 of 11

combinations thereof, on the probability of colonisation.The results are shown in Fig. 3.The dominant variable of the four appeared to be feeds

with breast milk, with increased feeds being associated withthe greatest shift towards low probability of C. perfringenscolonisation. We theorised two modes of action for this as-sociation, with breast milk acting either directly (inhibitingthe growth of C. perfringens) or indirectly (promoting thegrowth of other components of the gastro-intestinal micro-biota which outcompete C. perfringens). We explored thesepossibilities in culture experiments as follows.

Table 1 Demographics of infants included in the analysis (N =333)

Demographics Number

Gestation

Mean gestation at birth in days (SD) 197 (16)

Median gestation at birth in days (IQR) 198 (27)

Birth weight

Mean birth weight in g (SD) 1079.5(340.8)

Median birth weight in g (IQR) 1025 (525)

Gender

Female (%) 155 (46.5%)

Male (%) 178 (54.5%)

Ethnicity

White (%) 130 (39.0%)

Mixed (%) 44 (13.2%)

Asian (%) 58 (17.4%)

Black (%) 74 (22.2%)

Unknown (%) 12 (3.6%)

Other (%) 15 (4.5%)

Mode of delivery

Vaginal delivery (%) 137 (41.1%)

C-section (%) 196 (58.9%)

Ventilation

Mean number of days requiring mechanicalventilation prior to CPC or LS (SD)

3.1 (8.2)

Median number of days requiring mechanicalventilation prior to CPC or LS (IQR)

1 (2)

Mean number of days CPAP (air) prior to CPCor LS (SD)

7.0 (8.7)

Median number of days CPAP (air) prior toCPC or LS (IQR)

4 (10)

Mean number of days CPAP (oxygen) prior toCPC or LS (SD)

9.3 (14.8)

Median number of days CPAP (oxygen) priorto CPC or LS (IQR)

2 (12)

Feeding

Mean number of days of donor breast milkprior to CPC or LS (SD)

8.3 (8.9)

Median number of days of donor breast milkprior to CPC or LS (IQR)

6 (8)

Mean number of days of maternal breast milkprior to CPC or LS (SD)

23.0 (22.2)

Median number of days of maternal breast milkprior to CPC or LS (IQR)

17 (25)

Mean number of days of formula prior to CPCor LS (SD)

3.3 (9.0)

Median number of days of formula prior to CPC or LS(IQR)

0 (2)

Mean number of days of breast feeding prior to CPC orLS (SD)

3.4 (7.4)

Table 1 Demographics of infants included in the analysis (N =333) (Continued)

Demographics Number

Median number of days of breast feeding priorto CPC or LS (IQR)

0 (3)

Antibiotic use

Mean number of days of antibiotic use atbirth (SD)

2.4 (2.1)

Median number of days of antibiotic use atbirth (IQR)

2 (2)

Mean number of days of antibiotic use priorto CPC or LS (SD)

5.1 (7.4)

Median number of days of antibiotic use priorto CPC or LS (IQR)

3 (4)

C. perfringens colonisation

Number colonised with C. perfringens (%) 98 (29.4%)

Mean number of days prior to CPC (SD) 27.1 (23.3)

Median number of days prior to CPC (IQR) 21 (26)

Abbreviations: CPC C. perfringens colonisation, LS Last sample, CPAP,Continuous positive airway pressure, SD Standard deviation, IQRInterquartile range

Fig. 1 Kaplan-Meier plot of probability of colonisation by C.perfringens over time. Data from our cohort of 333 infants. Dashedlines indicate the Hall-Wellner 95% confidence bands

Shaw et al. BMC Pediatrics (2020) 20:75 Page 4 of 11

Growth of C. perfringens in breast milkA C. perfringens isolate was grown in either nutrient richmedium (supplemented WCB broth) or breast milk, eitheras a monoculture or in co-culture with Bifidobacteriuminfantis which was chosen to represent a typical competinggut species, one that thrives on breast milk oligosaccharides

[23–25]. Both species grew in each substrate and underboth culture conditions, and each grew significantly betterin rich medium under monoculture than in breast milk (B.infantis, p = 0.008, C. perfringens, p < 0.0001). There was nosignificant difference in growth between the two species inthe rich media when comparing the fold change between

Table 2 Results of the univariate survival analysis

Variable Quartiles CorrectedP value

Coefficient ExponentiatedCoefficient (95% CI)

Risk changefor 75%Quartile

0% 25% 50% 75% 100%

Gestation (days) 161 184 198 211 223 0.002 0.026 1.026 (1.013–1.040) 3.688

Birth weight (g) 500 800 1025 1325 1890 < 0.001 0.001 1.001 (1.001–1.002) 2.788

Days of mechanical ventilation 0 0 1 2 84 0.021 −0.076 0.927 (0.885–0.971) 0.860

Days of CPAP oxygen 0 0 2 12 80 < 0.001 −0.039 0.962 (0.947–0.977) 0.626

Days of maternal milk feeds(non-breast)

0 6 14 29 114 < 0.001 −0.057 0.945 (0.930–0.959) 0.192

Days of breast feeds 0 0 0 3 48 0.001 −0.073 0.929 (0.896–0.964) 0.802

Days of antibiotic use 0 2 3 6 79 < 0.001 −0.116 0.89 (0.848–0.934) 0.497

Fig. 2 Kaplan-Meier plots for each of the four significant factors in the multivariate model. X axis shows the infant day of life. Y axis shows theprobability of colonisation for an infant when stratified according to quartiles (1st – 4th) of varying clinical factors: a) Days of CPAP oxygen, b)Days of maternal milk feeds (excluding breast feeds), c) Days of breast feeding and d) Days of antibiotics usage. Colour codes for the quartiles areshown in the top right of each subplot

Shaw et al. BMC Pediatrics (2020) 20:75 Page 5 of 11

inoculation and the 20-h timepoint. However, in breastmilk C. perfringens grew at a significantly lower rate than B.infantis in monoculture (p = 0.003) and co-culture (p =0.0001). B. Infantis growth in breast milk was unaffected bymono- or co-culture, whilst growth of C. perfringens variedgreatly but without significant association to breast milktype or mono- or co-culture (Fig. 4).

Longitudinal carriage of C. perfringensWe sought to establish whether mother’s milk feeds orantibiotic treatments after the initial colonisation af-fected carriage of C. perfringens. 68 infants had at leastone additional sample available after C. perfringens wasfirst identified. Of these infants, 28% (n = 19) yielded nofurther isolates. For 31% of infants (n = 21), all subse-quent samples were positive, while the remaining 41% ofinfants (n = 28) tested positive in at least half of theiradditional samples. We recorded whether the infantsrecevied maternal milk feeds, CPAP oxygen or antibiotictreatment in the week after their initial colonisation, andfound none of the factors to be significantly asscoatedwith multiple isolation events compared to a single isola-tion (p = 0.75, 0.88 and 0.80 respectively).

Toxic potential of C. perfringens isolatesThe harmful effects of C. perfringens arise in large partthrough the production of toxins. Of particular interestin the neonatal field is the potential for C. perfringenstoxin to play a part in the pathogenesis of NEC [4, 5].We surveyed the C. perfringens isolates derived from ourcohort for the presence of toxin genes through targetedPCR and found that a range of toxin genes were present.The presence of toxin genes in C. perfringens isolatesduring any point in their admission was scored for eachneonate, allowing comparison of the toxic potential ofthe isolates prior to either discharge from the neonatalintensive care unit (“Control infants”) or NEC incidence(“NEC Infants”) (see Fig. 5).Statistical analyses (Barnard’s test and survival ana-

lysis) found no significant associations between the pres-ence of toxin genes and the development of NEC. Giventhe infant numbers available in our cohort, Barnard’stest would detect a proportional increase of 0.19 orgreater in the occurrence of toxin in infants developingNEC compared to controls with 95% confidence (assum-ing a one-sided test and 80% power). These resultstherefore do not support the hypothesis of NEC being

Fig. 3 Kaplan-Meier plots showing combinations of the four significant factors when split into low or high categories. For ease of comparability,‘All variables low’ (<= median value for all four variables) and ‘All variables high’ (>median value for all four variables) are shown on each chart

Shaw et al. BMC Pediatrics (2020) 20:75 Page 6 of 11

associated with the prevalence of a particular toxin geneacross our infant cohort. Overall, the most prevalenttoxin genes in our neonatal C. perfringens isolates werecpa, pfoA, colA and cpb2, which were found in 100, 64,99 and 50% of isolates respectively.

DiscussionWe have found that just under a third of infants (29.4%)in our premature neonatal cohort were colonised withC. perfringens in their gut at some time during their stayin the neonatal intensive care unit (NICU), with durationof maternal milk feeds, antibiotic therapy, and continu-ous positive airway pressure with supplemental oxygen(CPAP oxygen) treatment exerting the strongest influ-ence over probability of carriage. Our reported risk ofcolonisation with C. perfringens over time (approxi-mately 25% at 3 weeks and 44% at 3 weeks) is very simi-lar to findings in some other culture-dependent studies(35% at 3 weeks of life [14], 46.1% at 7 weeks of life[16]), although we are aware of one report of higherrates early in life (56.5% at 1 week of life [26]).We have found a significant inverse association between

both duration of maternal milk feeds and breast feedingand probability of C. perfringens colonisation. This is inline with studies of term infants, which have reported thatformula-fed infants have higher gut Clostridia counts

compared to those who were breast-fed [27–29]. Bioactiveconstituents (immunoglobulins, lysozyme, lactoferrin,antimicrobial peptides, oligosaccharides) and commensalmicrobes in breast milk [30] actively protect againstpathogen colonisation and invasion in the neonatal gut.An important mechanism of action is efficient metabolismof human milk oligosaccharides (HMOs) and subsequentacid production, for which B. infantis is especially adapted[23]. Many other gut commensal species, includingClostridium species [24, 25, 31], cannot metaboliseHMOs. Our growth experiments have shown that B. infan-tis significantly outperforms C. perfringens when grown inmonoculture (p = 0.003) and co-culture (p = 0.0001) inbreast milk. When the two species were cultured together,C. perfringens counts were not significantly reduced in thepresence of B. infantis (p = 0.1083) hence we saw no evi-dence of direct inhibitory action. This comparison could befurther confounded by a range of other factors; potentialcompetition by low abundance organisms derived from thebreast milk (predominantly Staphylococci), the impact of C.perfringen toxins on the growth of Bifidobacteria and the ef-fect of potential oxygen exposure which would impact C.perfringens more strongly than Bifidobacterium [32]. Directinhibition by specific Bifidobacterium species has howeverbeen previously reported for growth of enteric pathogens(including C. perfringens) [33, 34] in vitro; for growth of

Fig. 4 C. perfringens and B. infantis growth in rich media and breast milk. Each bacterium was grown in each medium separately (monoculture)or together (co-culture). Experiments were performed in triplicate with three technical replicates (all replicas shown). BM = Breast milk, with threedifferent donations being used in the experiments (a, b and c)

Shaw et al. BMC Pediatrics (2020) 20:75 Page 7 of 11

NEC-associated clostridia in a quail model [35]; andfor clostridial growth in mouse models [36]. It has re-cently been reported that term infants who are car-riers of C. perfringens have consistently lower levelsof Bifidobacteria in their gut compared to non-carriers [37].Interestingly, the impact of breast milk on C. perfringens

carriage we report was only found for infants fed their ownmother’s milk (breast-fed or expressed into bottles), andnot for those fed donor milk. In a recent systematic review[38], heat-treating donor milk to meet safety standards(normally by Holder pasteurization - 62.5 °C for 30min,[21]), was shown consistently to reduce the level and/or ac-tivity of milk proteins including immunoglobulins, lactofer-rin and enzymes, but had no effect on the lipid orsaccharide content of milk. This suggests that donor milkmay be inferior to maternal milk in an immunological cap-acity only, retaining qualities sufficient to influence C. per-fringens colonisation patterns in some contexts. It hasrecently been demonstrated that neonatal mice can be pro-tected from enteric infection by antibodies delivered

through breast milk [39]. The absence of competing milkcommensals in donor milk may perhaps further explainwhy in our study it failed to influence C. perfringens car-riage in the way that maternal milk did.Our results indicate that prolonged treatment with anti-

biotics or CPAP oxygen creates a hostile environment forC. perfringens in the preterm gut, reducing probability ofcarriage. C. perfringens may be particularly impacted byantibiotics due a rarity of multiple drug resistance genes[40] compared to other members of the neonatal gutmicrobiota [41]. A reduction in the abundance of C. per-fringens in the preterm infant gut with antibiotic exposurehas been observed in another cohort, although not re-duced prevalence [16]. While antibiotic prophylaxis is rou-tine in preterm neonatal care, it has been shown tosignificantly alter the gut microbiota in preterm infants[42, 43] and to be associated with adverse health outcomesincluding NEC, sepsis and death [44, 45]; the benefits ofavoiding putative C. perfringens-induced pathologies suchas NEC must be weighed against the risks of inducing gutdysbiosis when considering antibiotic regimes.

Fig. 5 C. perfringens toxin genes found in infants that developed NEC compared to control infants. Percentages were calculated out of the totalnumber of control infants (n = 91) and infants that developed NEC Bell stage 2 (confirmed) or 3 (severe) (n = 5). Two infants who developed NECBell stage 1 (suspected) could not be categorised as cases or controls so were not included. A toxin gene was scored as present if it was foundin any C. perfringens isolate found in an infant’s faecal samples during the course of either their whole time on the neonatal unit or up to the lastsample prior to NEC development. The table shows the counts within the two groups and the relative percentages for toxin occurrence

Shaw et al. BMC Pediatrics (2020) 20:75 Page 8 of 11

Similar to Ferraris et al., [16] we found that deliverymode had no effect on C. perfringens colonisation ratesin our preterm neonatal cohort, in contrast to term in-fants [46–48].Our analysis of the longitudinal carriage of C. perfringens

found no association between maternal milk feeds, CPAPoxygen or antibiotic treatment in the week after the initialisolation of C. perfringens and continued carriage (definedas at least half of subsequent samples testing positive). Ourstudy may however be underpowered to conclude whetherthese factors truly influence long term carriage as there waslimited time for observation between colonisation and theinfants being discharged from the NICU.Our toxin typing results in a limited number of C. per-

fringens-associated NEC cases did not identify a specifictoxin-mediated pathology. They have however providedevidence of a high occurrence of multi-toxigenic C. per-fringens strains in the premature neonatal gut. The threemost prevalent toxin genes in our neonatal C. perfrin-gens isolates (cpa, pfoA and colA) can all be located tothe same extracellular toxin gene cluster within a 250 kbregion on the chromosome [49]. Of note was the pres-ence of the plasmid-encoded beta2 toxin gene, whichwas found in50% of C. perfringens-positive infants. Al-though not fully elucidated, this toxin is strongly associ-ated with porcine necrotic enteritis and is cytotoxic forhuman colorectal epithelial (CaCo-2) cells [50]. Carriageof toxin genes, including beta2, is however observed inhealthy human populations [51], hence the toxic pheno-type is likely highly situational.We detected other plasmid-encoded toxin genes in

our neonatal isolates that have previously almost exclu-sively been detected in C. perfringens isolates from non-human species or cases of specific diseases: for example,the netB gene and necrotic enteritis in poultry, andbecAB genes and food poisoning in Japan. The most not-able example of this however, is the presence of the betatoxin gene (denoting type C C. perfringens when foundwith the alpha toxin and no other major toxin) in 17 in-fants. Type C C. perfringens in humans is normally onlyisolated in cases of necrotic enteritis (Pigbel) or fromhealthy humans in endemic areas. We hypothesise thatthere is a low level circulation of “aberrant” toxin genesin the neonatal population, presumably acquired fromthe mother at birth, and detectable in this study due tolarge sample sizes (273 isolates screened). It is importantto consider however, the high level of discordance be-tween genotype and phenotype for many of the C. per-fringens toxins.

ConclusionsWe sought to understand the C. perfringens colonisationdynamics in a cohort of infants at high risk of mortalityand morbidity from clostridia GI-induced pathologies.

We used a traditional culture method combined withPCR to demonstrate that toxigenic C. perfringens is partof the normal gut microbiota in preterm neonates andreport that approximately 30% are colonised before theyleave the NICU. We have identified perinatal factors thatare able to significantly affect the probability of C. per-fringens carriage: increased duration of maternal milkfeeds, increased duration of CPAP oxygen treatment,and increased duration of antibiotic treatment, whichcorrelated with protection from C. perfringens colonisa-tion. We demonstrated an inhibitory effect of breastmilk on the growth of C. perfringens in vitro and re-affirm the importance of maternal milk feeding in pre-term neonatal care.

Supplementary informationSupplementary information accompanies this paper at https://doi.org/10.1186/s12887-020-1976-7.

Additional file 1. C. perfringens toxin genes screened for in this study.

Additional file 2. PCR methods for toxin gene detection.

Additional file 3. Organisms grown in breast milk screening

AbbreviationsBHI: Brain heart infusion; BSM: Bifidus selective medium; CFU: Colony formingunits; CI: Confidence interval; CPAP: Continuous positive airway pressure;CPC: C. perfringens colonisation; CPE: C. perfringens enterotoxin; HMO: Humanmilk oligosaccharides; IQR: Interquartile range; LS: Last sample; MALDI-TOF: Matrix-assisted laser desorption/ ionization–time of flight;NEC: Necrotising enterocolitis; NICU: Neonatal intensive care unit;PCR: Polymerase chain reaction; SD: Standard deviation; TSC: Tryptosesulphite cycloserine; WCB: Wilkins-chalgren broth

AcknowledgementsWe thank the microbiology laboratory at Charing Cross Hospital, ImperialCollege Healthcare NHS Trust London for use of the MALDI-TOF facilities. Wethank the participants and their families for their contribution to the study.

Authors’ contributionsAGS contributed to the manuscript, designed and conducted culture-basedexperimental work and performed the statistical analyses. EC contributed tothe manuscript, designed the competition experiments and conducted theculture-based experimental work. KS contributed to the manuscript and theculture-based experimental work. HS and HT conducted the culture-basedexperimental work and assembled cohort data. JCSB contributed to themanuscript and designed and conducted the toxin PCR experiments. RABand JSK contributed to the experimental design. All authors read and ap-proved the final manuscript.

FundingThis work was supported by funding from the Winnicott Foundation;Micropathology Ltd.; Innovate UK; and the National Institute for HealthResearch (NIHR) Biomedical Research Centre based at Imperial HealthcareNHS Trust and Imperial College London. KS was funded during this work byan NIHR Doctoral Research Fellowship [NIHR-DRF-2011-04-128]. This articlepresents independent research funded by the NIHR. The views expressed arethose of the authors and not necessarily those of the NHS, the NIHR, theDepartment of Health or other funders. Funding bodies had no role in thedesign of the study and collection, analysis, and interpretation of data and inwriting the manuscript.

Availability of data and materialsNot applicable.

Shaw et al. BMC Pediatrics (2020) 20:75 Page 9 of 11

Ethics approval and consent to participateThe study “Defining the Intestinal Microbiota in Premature Infants” (ClinicalTrials.govidentifier NCT01102738) was approved by West London Research Ethics CommitteeTwo (National Health Service Health Research Authority), United Kingdom (referencenumber 10/H0711/39). Parents gave written informed consent for their infant toparticipate in the study.

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests.

Author details1Department of Infectious Disease Epidemiology, Imperial College London,London, UK. 2Department of Medicine, Section of Paediatrics, ImperialCollege London, London, UK. 3School of Life Sciences, University of Lincoln,Lincoln, UK.

Received: 1 July 2019 Accepted: 12 February 2020

References1. Stevens DL, Aldape MJ, Bryant AE. Life-threatening clostridial infections.

Anaerobe. 2012;18(2):254–9.2. Centers for Disease Control and Prevention. Food Safety Homepage,

Foodborne Illness A-Z, Clostridium perfringens. http://www.cdc.gov/foodsafety/diseases/clostridium-perfringens.html 2015. Accessed Nov 2017.

3. Neu J, Walker WA. Necrotizing enterocolitis. N Engl J Med. 2011;364(3):255–64.

4. Blakey JL, Lubitz L, Campbell NT, Gillam GL, Bishop RF, Barnes GL. Entericcolonization in sporadic neonatal necrotizing enterocolitis. J PediatrGastroenterol Nutr. 1985;4(4):591–5.

5. Dittmar E, Beyer P, Fischer D, Schäfer V, Schoepe H, Bauer K, et al.Necrotizing enterocolitis of the neonate with Clostridium perfringens:diagnosis, clinical course, and role of alpha toxin. Eur J Pediatr. 2008;167(8):891–5.

6. Heida FH, van Zoonen AG, Hulscher JB, te Kiefte BJ, Wessels R, Kooi EM,et al. A necrotizing Enterocolitis-associated gut microbiota is present in themeconium: results of a prospective study. Clin Infec Dis : Official PublicationInfec Dis Soc Am. 2016;62(7):863–70.

7. Sim K, Shaw AG, Randell P, Cox MJ, McClure ZE, Li M-S, et al. Dysbiosisanticipating necrotizing Enterocolitis in very premature infants. Clin InfecDis: Official Publication Infec Dis Soc Am. 2015;60(3):389–97.

8. Warner BB, Deych E, Zhou Y, Hall-Moore C, Weinstock GM, Sodergren E,et al. Gut bacteria dysbiosis and necrotising enterocolitis in very lowbirthweight infants: a prospective case-control study. Lancet. 2016;387(10031):1928–36.

9. Dobbler PT, Procianoy RS, Mai V, Silveira RC, Corso AL, Rojas BS, et al. Lowmicrobial diversity and abnormal microbial succession is associated withnecrotizing Enterocolitis in preterm infants. Front Microbiol. 2017;8:2243.

10. Penders J, Thijs C, van den Brandt PA, Kummeling I, Snijders B, Stelma F,et al. Gut microbiota composition and development of atopicmanifestations in infancy: the KOALA birth cohort study. Gut. 2007;56:661–7.

11. Björkstén B, Sepp E, Julge K, Voor T, Mikelsaar M. Allergy development andthe intestinal microflora during the first year of life. J Allergy Clin Immunol.2001;108(4):516–20.

12. Martirosian G, Ekiel A, Aptekorz M, Wiechula B, Kazek B, Jankowska-Steifer E,et al. Fecal lactoferrin and Clostridium spp. in stools of autistic children.Anaerobe. 2011;17(1):43–5.

13. Finegold SM, Molitoris D, Song Y, Liu C, Vaisanen ML, Bolte E, et al.Gastrointestinal microflora studies in late-onset autism. Clin Infect Dis. 2002;35(Suppl 1):S6–s16.

14. Blakey JL, Lubitz L, Barnes GL, Bishop RF, Campbell NT, Gillam GL.Development of gut colonisation in pre-term neonates. J Med Microbiol.1982;15(4):519–29.

15. Arboleya S, Binetti A, Salazar N, Fernández N, Solís G, Hernández-Barranco A,et al. Establishment and development of intestinal microbiota in pretermneonates. FEMS Microbiol Ecol. 2012;79(3):763–72.

16. Ferraris L, Butel MJ, Campeotto F, Vodovar M, Rozé JC, Aires J. Clostridia inpremature Neonates' gut: incidence, antibiotic susceptibility, and perinataldeterminants influencing colonization. PLoS One. 2012;7(1):e30594.

17. Kiu R, Hall LJ. An update on the human and animal enteric pathogenClostridium perfringens. Emerg Microbes Infect. 2018;7(1):141.

18. Uzal FA, Freedman JC, Shrestha A, Theoret JR, Garcia J, Awad MM, et al.Towards an understanding of the role of Clostridium perfringens toxins inhuman and animal disease. Future Microbiol. 2014;9(3):361–77.

19. van Asten AJAM, van der Wiel CW, Nikolaou G, Houwers DJ, Gröne A. Amultiplex PCR for toxin typing of Clostridium perfringens isolates. VetMicrobiol. 2009;136(3):411–2.

20. Public Health England. UK standards for microbiology investigations: processingof faeces for Clostridium difficile.: Available at: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/343912/B_10i1.5.pdf; 2014.Accessed Nov 2017.

21. Excellence NIfCaH. Donor milk banks: service operation. In: NICE GuidanceCgC, editor.: Available at https://www.nice.org.uk/guidance/CG93/chapter/1-Guidance; 2010. Accessed Nov 2017.

22. Kotsanas D, Carson JA, Awad MM, Lyras D, Rood JI, Jenkin GA, et al. Noveluse of Tryptose sulfite Cycloserine egg yolk agar for isolation of Clostridiumperfringens during an outbreak of necrotizing Enterocolitis in a neonatalunit. J Clin Microbiol. 2010;48(11):4263–5.

23. Sela DA, Chapman J, Adeuya A, Kim JH, Chen F, Whitehead TR, et al. Thegenome sequence of Bifidobacterium longum subsp. infantis revealsadaptations for milk utilization within the infant microbiome. Proc NatlAcad Sci U S A. 2008;105(48):18964–9.

24. Marcobal A, Barboza M, Froehlich JW, Block DE, German JB, Lebrilla CB, et al.Consumption of human Milk oligosaccharides by gut-related microbes. JAgric Food Chem. 2010;58(9):5334–40.

25. Rockova S, Rada V, Marsik P, Vlkova E, Bunesova V, Sklenar J, et al. Growth ofbifidobacteria and clostridia on human and cow milk saccharides.Anaerobe. 2011;17(5):223–5.

26. Rotimi VO, Olowe SA, Ahmed I. The development of bacterial flora ofpremature neonates. J Hyg. 1985;94(3):309–18.

27. Stark PL, Lee A. The microbial ecology of the large bowel of breast-fed andformula-fed infants during the first year of life. J Med Microbiol. 1982;15(2):189–203.

28. Benno Y, Sawada K, Mitsuoka T. The intestinal microflora of infants:composition of fecal flora in breast-fed and bottle-fed infants. MicrobiolImmunol. 1984;28(9):975–86.

29. Fallani M, Amarri S, Uusijarvi A, Adam R, Khanna S, Aguilera M, et al.Determinants of the human infant intestinal microbiota after theintroduction of first complementary foods in infant samples from fiveEuropean centres. Microbiology (Reading, England). 2011;157(Pt 5):1385–92.

30. Fitzstevens JL, Smith KC, Hagadorn JI, Caimano MJ, Matson AP, Brownell EA.Systematic review of the human Milk microbiota. Nutr Clin Pract : OfficialPublication Am Soc Parenter Enteral Nutr. 2017;32:354–64.

31. Yu ZT, Chen C, Kling DE, Liu B, McCoy JM, Merighi M, et al. The principalfucosylated oligosaccharides of human milk exhibit prebiotic properties oncultured infant microbiota. Glycobiology. 2013;23(2):169–77.

32. Rolfe RD, Hentges DJ, Campbell BJ, Barrett JT. Factors related to the oxygentolerance of anaerobic bacteria. Appl Environ Microbiol. 1978;36(2):306–13.

33. Hutt P, Shchepetova J, Loivukene K, Kullisaar T, Mikelsaar M. Antagonisticactivity of probiotic lactobacilli and bifidobacteria against entero- anduropathogens. J Appl Microbiol. 2006;100(6):1324–32.

34. Martinez FA, Balciunas EM, Converti A, Cotter PD, de Souza Oliveira RP. Bacteriocinproduction by Bifidobacterium spp. Rev Biotechnol Adv. 2013;31(4):482–8.

35. Butel MJ, Roland N, Hibert A, Popot F, Favre A, Tessedre AC, et al. Clostridialpathogenicity in experimental necrotising enterocolitis in gnotobiotic quailsand protective role of bifidobacteria. J Med Microbiol. 1998;47(5):391–9.

36. Fukuda S, Toh H, Hase K, Oshima K, Nakanishi Y, Yoshimura K, et al.Bifidobacteria can protect from enteropathogenic infection throughproduction of acetate. Nature. 2011;469(7331):543–7.

37. Nagpal R, Tsuji H, Takahashi T, Nomoto K, Kawashima K, Nagata S, et al. Gutdysbiosis following C-section instigates higher colonisation of toxigenicClostridium perfringens in infants. Benefic Microbes. 2017;8(3):353–65.

38. Peila C, Moro GE, Bertino E, Cavallarin L, Giribaldi M, Giuliani F, et al. TheEffect of Holder Pasteurization on Nutrients and Biologically-ActiveComponents in Donor Human Milk: A Review. Nutrients. 2016;8(8):477.https://doi.org/10.3390/nu8080477.

Shaw et al. BMC Pediatrics (2020) 20:75 Page 10 of 11

39. Zheng W, Zhao W, Wu M, Song X, Caro F, Sun X, et al. Microbiota-targetedmaternal antibodies protect neonates from enteric infection. Nature. 2020;577(7791):543–8.

40. Kiu R, Caim S, Alexander S, Pachori P, Hall LJ. Probing genomic aspects ofthe multi-host pathogen Clostridium perfringens reveals significantPangenome diversity, and a diverse Array of virulence factors. FrontMicrobiol. 2017;8:2485.

41. Rose G, Shaw AG, Sim K, Wooldridge DJ, Li MS, Gharbia S, et al. Antibioticresistance potential of the healthy preterm infant gut microbiome. PeerJ.2017;5:e2928.

42. Arboleya S, Sánchez B, Milani C, Duranti S, Solís G, Fernández N, et al.Intestinal microbiota development in preterm neonates and effect ofperinatal antibiotics. J Pediatr. 2015;166(3):538–44.

43. Greenwood C, Morrow AL, Lagomarcino AJ, Altaye M, Taft DH, Yu Z, et al.Early empiric antibiotic use in preterm infants is associated with lowerbacterial diversity and higher relative abundance of Enterobacter. J Pediatr.2014;165(1):23–9.

44. Kuppala VS, Meinzen-Derr J, Morrow AL, Schibler KR. Prolonged initialempirical antibiotic treatment is associated with adverse outcomes inpremature infants. J Pediatr. 2011;159(5):720–5.

45. Cotten CM, Taylor S, Stoll B, Goldberg RN, Hansen NI, Sanchez PJ, et al.Prolonged duration of initial empirical antibiotic treatment is associatedwith increased rates of necrotizing enterocolitis and death for extremelylow birth weight infants. Pediatrics. 2009;123(1):58–66.

46. Penders J, Thijs C, Vink C, Stelma FF, Snijders B, Kummeling I, et al. Factorsinfluencing the composition of the intestinal microbiota in early infancy.Pediatrics. 2006;118(2):511–21.

47. Gronlund MM, Lehtonen OP, Eerola E, Kero P. Fecal microflora in healthyinfants born by different methods of delivery: permanent changes inintestinal flora after cesarean delivery. J Pediatr Gastroenterol Nutr. 1999;28(1):19–25.

48. Biasucci G, Rubini M, Riboni S, Morelli L, Bessi E, Retetangos C. Mode ofdelivery affects the bacterial community in the newborn gut. Early HumDev. 2010;86(Suppl 1):13–5.

49. Katayama S, Dupuy B, Garnier T, Cole ST. Rapid expansion of the physicaland genetic map of the chromosome of Clostridium perfringens CPN50. JBacteriol. 1995;177(19):5680–5.

50. Fisher DJ, Miyamoto K, Harrison B, Akimoto S, Sarker MR, McClane BA.Association of beta2 toxin production with Clostridium perfringens type ahuman gastrointestinal disease isolates carrying a plasmid enterotoxin gene.Mol Microbiol. 2005;56(3):747–62.

51. Lakshminarayanan B, Harris HM, Coakley M, O'Sullivan O, Stanton C,Pruteanu M, et al. Prevalence and characterization of Clostridiumperfringens from the faecal microbiota of elderly Irish subjects. J MedMicrobiol. 2013;62(Pt 3):457–66.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Shaw et al. BMC Pediatrics (2020) 20:75 Page 11 of 11